| Abstract | Information sharing on online travel sites in the form of online reviews has emerged as
the soft component in service-dominant destination management. Thus, text analytics of online reviews can make inferences based on thousands of tourists’ opinions,
feedback, and discussion on online travel sites to make the destination management
process smart and intelligent. Hence, the main objective of this research study is to
extract the underlying topics and sentiment polarity expressed in the online reviews.
This research study was carried out scraping 44,161 English reviews of attraction of
Nepal including Sights and Landmarks, Nature and Parks, and Museum from
TripAdvisor. The topic modeling algorithm (Latent Dirichlet Allocation) identified six topics for Sights and Landmarks Attraction (Ambience, Cost, Route and Transportation, Cultural Traditions, Infrastructure Conservation, and Shopping Market), four topics for Nature and Parks Attraction (Sightseeing, Travel Activities, Ambience, and Cost) and lastly
four topics for Museum Attraction (Exhibition, Building/Infrastructure, Ticketing and Facilities, and Mountaineering Information and Collections). Lexicon-based sentiment analysis using VADER was performed to categorize reviews into positive, negative,
and neutral classes. The sentiment expressed in topics under all categories were mostly
positive. Furthermore, the negative reviews under each topic identified were analyzed
to understand the existing problems that tourists mostly complain about. This study
finds that the sentiment score calculated by VADER has a moderate positive
relationship with the bubble rating provided by the TripAdvisor users. The result
suggests that there is a disparity between neutral and negative sentiment scores assigned
by TripAdvisor users and VADER. By analyzing the reviews, it was observed that TripAdvisor users tend to give a higher score to negative reviews. |